Model training method, fault diagnosis method, equipment, medium and product

Abstract

The invention relates to the technical field of aero-engines, and particularly provides a model training method, a fault diagnosis method, equipment, a medium and a product. The method comprises the following steps: acquiring historical fault case data in a fault case library, wherein the historical fault case data comprises fault description data and a fault solution; obtaining a fault prediction solution based on the fault description data and a neural network model to be trained; calculating model loss based on the fault solution and the fault prediction solution; and optimizing parameters of the neural network model by adopting the model loss until the neural network model converges. In this way, when a fault needs to be diagnosed, the converged neural network model can be adopted to diagnose the fault to be diagnosed, and intelligent diagnosis of the fault to be diagnosed is achieved. Compared with instrument diagnosis or experience diagnosis in the prior art, the fault diagnosis efficiency can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of aviation engine technology, and in particular to a model training method, a fault diagnosis method, a device, a medium and a product. Background Art

[0002] The aircraft engine production process involves a large number of mechanical, hydraulic, electrical and other parts, and their failure forms and mechanisms are different. Faced with various forms of failure, traditional subjective diagnosis, instrument diagnosis and other methods are still used in most cases. There are great individual differences in subjective diagnosis, which belongs to empirical judgment and has poor accuracy. Instrument diagnosis is expensive, time-consuming and labor-intensive. Therefore, existing methods cannot troubleshoot efficiently and quickly, and can hardly adapt to the maintenance rhythm in combat scenarios. Summary of the invention

[0003] In view of the above problems, the present disclosure is proposed. The present disclosure provides a model training method, a fault diagnosis method, a device, a medium and a product.

[0004] According to a first aspect of the present disclosure, a model training method is provided, comprising: Acquire historical fault case data in a fault case library, wherein the historical fault case data includes fault description data and fault solutions; Obtaining a fault prediction solution based on the fault description data and the neural network model to be trained; calculating a model loss based on the fault solution and the fault prediction solution; The model loss is used to optimize the parameters of the neural network model until the neural network model converges.

[0005] In addition, according to the model training method of the first aspect of the present disclosure, a fault prediction solution is obtained based on the fault description data and the neural network model to be trained, including: Extracting keywords from the fault description data according to predefined fault feature attributes to obtain fault feature keywords; Generating a fault feature vector of the fault feature keyword; The fault feature vector is input into the neural network model to obtain the fault prediction solution.

[0006] In addition, according to the model training method of the first aspect of the present disclosure, the fault characteristic attribute includes at least one of the following: Operating environment; The fault belongs to; Fault phenomenon.

[0007] In addition, the model training method according to the first aspect of the present disclosure further includes: Obtaining a generation system type of the historical fault case data; The historical fault case data is stored in the fault case library according to the generation system type.

[0008] According to a second aspect of the present disclosure, a fault diagnosis method is provided, comprising: Obtain target fault description data of the fault to be diagnosed; Using the converged neural network model obtained by the model training method based on the first aspect and the target fault description data, predicting the current fault prediction solution of the fault to be diagnosed; Based on the current fault solution, obtaining a target historical fault diagnosis case having the highest similarity to the fault to be diagnosed from a historical fault diagnosis library; The fault diagnosis solution in the target historical fault diagnosis case is used as the fault solution for the fault to be diagnosed.

[0009] In addition, according to the fault diagnosis method of the second aspect of the present disclosure, based on the current fault solution, obtaining the target historical fault diagnosis case with the highest similarity to the fault to be diagnosed from the historical fault diagnosis library includes: Extracting keywords from the target fault description data according to predefined fault feature attributes to obtain target fault feature keywords; Generate a target fault feature vector of the target fault feature keyword; Clustering the target fault feature vector and the fault feature vectors corresponding to the historical fault case data of the fault case library to obtain similar cases having the same characteristic attributes as the fault to be diagnosed; The similarity between the current fault solution and each of the similar cases is calculated, and the historical fault diagnosis case with the highest similarity is used as the target historical fault diagnosis case.

[0010] In addition, according to the fault diagnosis method of the second aspect of the present disclosure, after taking the fault diagnosis solution in the target historical fault diagnosis case as the fault solution for the fault to be diagnosed, the method further includes: When it is determined that the fault description data of the target historical fault diagnosis case is not completely the same as the target fault description data, the fault to be diagnosed and the fault solution to the fault to be diagnosed are updated to the fault history diagnosis library.

[0011] According to a third aspect of the present disclosure, there is provided a fault diagnosis device, comprising: An acquisition module is used to acquire target fault description data of a fault to be diagnosed; A prediction module, configured to predict a current fault prediction solution for the fault to be diagnosed by using a converged neural network model obtained by the model training method according to the first aspect and the target fault description data; An acquisition module, used to acquire, from a historical fault diagnosis library, a target historical fault diagnosis case having the highest similarity to the fault to be diagnosed based on the current fault solution; A determination module is used to use the fault diagnosis solution in the target historical fault diagnosis case as the fault solution for the fault to be diagnosed.

[0012] According to a fourth aspect of the present disclosure, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the first aspect. According to a fifth aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program / instruction is stored, and when the computer program / instruction is executed by a processor, the steps of the method described in the first aspect are implemented. According to a sixth aspect of the present disclosure, there is provided a computer program product, comprising a computer program / instruction, which implements the steps of the method described in the first aspect when executed by a processor.

[0013] As will be described in detail below, according to the model training method and fault diagnosis method of the embodiment of the present disclosure, the historical fault data in the fault case library is used to train the neural network model, and a converged neural network model is obtained, so that when there is a fault to be diagnosed, the converged neural network model can be used to diagnose the fault to be diagnosed, thereby realizing intelligent diagnosis of the fault to be diagnosed. Compared with instrument diagnosis or empirical diagnosis in related technologies, the efficiency of fault diagnosis can be improved.

[0014] It is to be understood that both the foregoing general description and the following detailed description are exemplary, and are intended to provide further explanation of the technology as claimed. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The above and other purposes, features and advantages of the present disclosure will become more apparent by describing the embodiments of the present disclosure in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and do not constitute a limitation of the present disclosure. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 is a flowchart illustrating the application of a model training method according to an embodiment of the present disclosure.

[0017] Figure 2is a flowchart illustrating the application of the fault diagnosis method according to an embodiment of the present disclosure.

[0018] Figure 3 is a flowchart illustrating the application of a model training method according to an embodiment of the present disclosure.

[0019] Figure 4 is a structural diagram illustrating an application of a model training device according to an embodiment of the present disclosure.

[0020] Figure 5 is a structural diagram illustrating application of a fault diagnosis method according to an embodiment of the present disclosure.

[0021] Figure 6 is a hardware block diagram illustrating an electronic device according to an embodiment of the present disclosure.

[0022] Figure 7 is a schematic diagram illustrating a computer-readable storage medium according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical scheme and advantages of the embodiments of the present disclosure clearer, the technical scheme in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. The components of the embodiments of the present disclosure generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the drawings is not intended to limit the scope of the present disclosure for protection, but merely represents the selected embodiments of the present disclosure. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present disclosure.

[0024] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0025] The term "and / or" herein only describes an association relationship, indicating that three relationships may exist. For example, A and / or B may represent the following three situations: A exists alone, A and B exist at the same time, and B exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C may represent including any one or more elements selected from the set consisting of A, B, and C.

[0026] In related technologies, traditional subjective diagnosis or instrument diagnosis methods are used to diagnose the causes of aircraft engine failures. Both methods cannot efficiently and quickly troubleshoot, and are difficult to adapt to the maintenance rhythm in combat scenarios.

[0027] In order to alleviate the technical problems existing in the related art, the embodiments of the present disclosure provide a model training method, a fault diagnosis method, a device, a medium and a product. The method uses the historical fault data in the fault case library to train the neural network model and obtain a converged neural network model, so that when there is a fault to be diagnosed, the converged neural network model can be used to diagnose the fault to be diagnosed, thereby realizing intelligent diagnosis of the fault to be diagnosed. Compared with the instrument diagnosis or experience diagnosis in the related art, the efficiency of fault diagnosis can be improved.

[0028] To facilitate understanding of this embodiment, a model training method disclosed in the embodiment of the present disclosure is first introduced in detail. The execution subject of the model training method provided in the embodiment of the present disclosure is generally an electronic device with certain computing capabilities, and the electronic device includes, for example, a terminal device or a server or other processing device. In some possible implementations, the model training method can be implemented by a processor calling a computer-readable instruction stored in a memory.

[0029] See also Figure 1 FIG. 1 is a flowchart of a model training method provided by an embodiment of the present disclosure, and the method comprises the following steps: Step 101: Acquire historical fault case data in a fault case library, wherein the historical fault case data includes fault description data and fault solutions.

[0030] In this embodiment, the historical fault case data mainly comes from the collection and organization of troubleshooting and maintenance records during the engine production process, including fault repair manuals and maintenance service on-site work minutes.

[0031] The fault maintenance manual covers the fault description data and fault solutions of common engine faults provided by equipment parts manufacturers and users, including but not limited to fault phenomena, faulty equipment, fault causes and fault location, etc. The maintenance service on-site work minutes are the fault description data and fault solutions provided by equipment parts manufacturers and users when performing on-site maintenance for faults that have never occurred in the engine. The collected fault case data will be organized uniformly, low-value and redundant data items will be eliminated, and a fault case library will be obtained to improve the quality and integrity of the case data.

[0032] Step 102: Obtain a fault prediction solution based on the fault description data and the neural network model to be trained.

[0033] In this embodiment, the neural network model to be trained includes but is not limited to a radial basis function (RBF) neural network model.

[0034] In some embodiments, step 102 may include the following steps: According to the pre-defined fault feature attributes, keywords are extracted from the fault description data to obtain fault feature keywords; Generate a fault feature vector of the fault feature keyword; The fault feature vector is input into the neural network model to obtain the fault prediction solution.

[0035] The predefined fault characteristic attributes include, but are not limited to, operating environment, fault category and / or fault phenomenon.

[0036] The operating environment is used to characterize the environment in which the faulty component of the engine is located. The operating environment has multiple categories of definitions, including temperature, humidity, salinity, altitude, etc. Recording this information can provide a basis for quickly matching similar faults triggered in the same operating environment.

[0037] Fault attribution is represented in three levels: component level, part level, and subassembly level. Each fault must record the component, part, or subassembly to which it belongs. If it is a part or subassembly level fault, the lower level may not be recorded. For example, a hydraulic outrigger fault has a component that is a hydraulic cylinder seal, a part that is a hydraulic cylinder, and a subassembly that is a hydraulic outrigger. This representation method can quickly establish an association relationship for subsequent component faults of the same level or category.

[0038] Fault phenomenon, by extracting the characteristic words of the input continuous information data, obtain the phenomenon keywords of the fault. There is no limit on the number of this attribute. This attribute is the core attribute of this system and is the basis for analysis and reasoning. For example, a fault phenomenon can be described as unstable engine shaft speed.

[0039] In this embodiment, the fault keyword may be converted into a fault feature vector based on the TF-IDF value of the fault keyword.

[0040] Among them, the TF - IDF value of the fault keyword is:

[0041] in, is the number of times the fault keyword ω appears in this case, is the sum of the number of occurrences of all words in the case; is the number of cases containing the word ω, plus 1 to avoid the denominator being 0, is the total number of cases in the case library.

[0042] In this embodiment, after the fault feature words are converted into fault feature vectors, the fault feature vectors are used as inputs of the neural network model. In the neural network model, the fault feature vector X=[x1, x2, ...xn] is input, where X is generated by fault keywords in historical fault cases; the output vector is Y=[y1, y2, ...ym], i.e., the predicted fault solution.

[0043] In this embodiment, the hidden layer of the neural network model uses a radial basis function, and its weighted network output is:

[0044] in, is the output of the jth node in the output layer, is the output of the i-th node in the hidden layer, is the weight coefficient from the hidden layer to the output layer, , is the center and variance of the Gaussian function of the i-th node, ||·|| is the input x and The distance between them, n, h, m are the number of nodes in each layer. In the RBF network, the center of the hidden layer function ,variance and weighting coefficients , whose value is obtained through the network's learning and training process of the actual case library cases. In the unsupervised learning process, the center and variance of the hidden layer basis function are solved; in the supervised learning process, the weights between the hidden layer and the output layer are solved. The specific calculation process is: Select h centers for clustering. For the radial basis of the Gaussian kernel function, the variance is solved by the formula: ,(i=1,2, …h) in, is the maximum distance between the selected center points. Weighting coefficient It is directly calculated by the least squares method, that is, solving the loss function about The partial derivative of , so that it is equal to 0, is calculated as: ,( ) Step 103: Calculate model loss based on the fault solution and the fault prediction solution.

[0045] Step 104: Optimize the parameters of the neural network model using the model loss until the neural network model converges.

[0046] In some optional embodiments, the historical fault data in the historical fault case library can also be saved based on the generation system type. In specific implementation, the method can also include the following steps: The type of system that generates the historical failure case data; The historical fault case data is stored in the fault case library according to the generation system type.

[0047] The generated system type includes but is not limited to a mechanical system, an electronic system or a hydraulic system. The generated system type of the historical fault case data may be manually indicated.

[0048] In the solution provided by this embodiment, the historical fault data in the fault case library is used to train the neural network model, and a converged neural network model is obtained, so that when there is a fault to be diagnosed, the converged neural network model can be used to diagnose the fault to be diagnosed, thereby realizing intelligent diagnosis of the fault to be diagnosed. Compared with instrument diagnosis or experience diagnosis in related technologies, the efficiency of fault diagnosis can be improved.

[0049] The present disclosure also provides a fault diagnosis method, the execution subject of the fault diagnosis method is generally an electronic device with certain computing capabilities, such as a terminal device or a server or other processing device. In some possible implementations, the fault diagnosis method can be implemented by a processor calling a computer-readable instruction stored in a memory.

[0050] See also Figure 2 FIG. 1 is a flowchart of a fault diagnosis method provided by an embodiment of the present disclosure, and the method comprises the following steps: Step 201: Obtain target fault description data of a fault to be diagnosed.

[0051] Step 202: Use the converged neural network model obtained based on the model training method in the above-mentioned embodiment and the target fault description data to predict the current fault prediction solution for the fault to be diagnosed.

[0052] Step 203: Based on the current fault solution, obtain a target historical fault diagnosis case with the highest similarity to the fault to be diagnosed from a historical fault diagnosis library.

[0053] Step 204: Use the fault diagnosis solution in the target historical fault diagnosis case as the fault solution for the fault to be diagnosed.

[0054] In an optional embodiment, step 203 may include the following steps: According to the pre-defined fault feature attributes, keywords are extracted from the target fault description data to obtain target fault feature keywords; Generate a target fault feature vector of a target fault feature keyword; Clustering the target fault feature vector and the fault feature vectors corresponding to the historical fault case data in the fault case library to obtain similar cases with the same characteristic attributes as the fault to be diagnosed; Calculate the similarity between the current fault solution and each similar case, and use the historical fault diagnosis case with the highest similarity as the target historical fault diagnosis case.

[0055] It should be understood that the predefined fault characteristic attributes here are the same as the predefined fault attributes in the aforementioned model training method. Therefore, the predefined fault characteristic attributes in this embodiment also include but are not limited to the operating environment, fault affiliation and / or fault phenomenon.

[0056] The process of generating the target fault feature vector of the target fault feature keyword in this embodiment is the same as the process of generating the fault feature vector of the fault keyword in the embodiment of the aforementioned model training method, that is, converting the target fault keyword into the target fault feature vector based on the TF-IDF value of the target feature keyword.

[0057] In this embodiment, the K-means clustering algorithm is used to cluster the target fault feature vector and the fault feature vector corresponding to the historical fault case data in the fault case library. The basic idea of ​​the K-means clustering algorithm is to divide the data into K different clusters in an iterative manner, and minimize the sum of the distances between each data point and the centroid (or center point, mean point) of the cluster to which it belongs. Therefore, the K-means clustering algorithm can be used to find cases with the same category as the fault to be diagnosed, that is, similar cases, from the fault case library.

[0058] It should be understood that when clustering, in order to reduce the amount of calculation and the complexity of the calculation, it is also possible to first determine the generation system category of the fault to be diagnosed, and then find a historical fault case data set belonging to the generation system category from the fault case library, and use the K-means clustering algorithm to cluster the target fault feature vector and the fault feature vector corresponding to the historical fault case data in the historical fault case data set, so as to find similar cases with the same characteristic attributes as the fault to be diagnosed.

[0059] It should be understood that similar cases may include historical fault case data that is exactly the same as the case to be diagnosed. Therefore, in actual applications, in order to simplify calculations, before calculating the similarity between the current fault solution and each similar case, the user can also indicate through the human-computer interaction interface whether there are the same cases of the fault to be diagnosed in the same cases, and when the user indicates that there are no the same cases of the fault to be diagnosed in the same cases, the similarity between the current fault solution and each similar case is calculated. It should be understood that when there are the same cases of the case to be diagnosed in the same cases, the fault solution of the same case is directly used as the fault solution of the case to be diagnosed.

[0060] In some embodiments, after the fault diagnosis solution in the target historical fault diagnosis case is used as the fault solution for the fault to be diagnosed, the following steps may also be included: When it is determined that the fault description data of the target historical fault diagnosis case is not completely the same as the target fault description data, the fault to be diagnosed and the fault solution to the fault to be diagnosed are updated to the fault history diagnosis library.

[0061] The following is a flowchart of the fault diagnosis method combined with the model training process. Figure 3 As shown, the method may include the following steps: Step 301: Fault case data collection.

[0062] The historical fault case data used in this embodiment mainly comes from the collection and organization of troubleshooting and maintenance records during the engine production process, including fault maintenance manuals and maintenance service on-site work minutes. The fault maintenance manual covers the descriptions and solutions of common engine faults by equipment parts manufacturers and users, including fault phenomena, faulty equipment, fault causes, and fault location; the maintenance service on-site work minutes are fault descriptions and solutions issued by equipment parts manufacturers and users when performing on-site maintenance for faults that have not occurred in the engine. The collected fault data is uniformly organized, and low-value and redundant data items are eliminated to improve the quality and integrity of the case data.

[0063] Step 302: Feature extraction.

[0064] According to the standard definition of fault feature attributes, keywords are extracted from the fault description data of the historical fault case data, including but not limited to using the Jieba word segmentation method to segment the fault description data to extract fault keywords.

[0065] The fault feature attribute can be expressed as Pi = (Pi1, Pi2, Pi3, Pi4), (i = 1, 2, ..., n), where n is the number of fault cases in the historical fault case data. Specifically, Pi1 represents the operating environment, which is used to characterize the environment in which the faulty component in the engine is located. The operating environment has multiple categories of definitions, including temperature, humidity, salinity, altitude, etc. Recording this information can provide a basis for quickly matching similar faults triggered under the same operating environment. Pi2: Indicates the fault category, using a three-level representation, which can be expressed as component level, part level, and component level. Each fault needs to record the component, part, and component to which it belongs. If it is a part or component level fault, the lower level can be omitted. For example: a hydraulic outrigger fault, its component is a hydraulic cylinder seal, its part is a hydraulic cylinder, and its component is a hydraulic outrigger. This representation method can quickly establish an association relationship for subsequent faults of the same level and category. Pi3: By extracting the feature words of the input continuous information data, the phenomenon keywords of the fault are obtained. There is no limit on the number of this attribute. This attribute is the core attribute of this system and the basis for analysis and reasoning. For example, a fault phenomenon can be described as unstable engine shaft speed.

[0066] Step 303: Enter the fault case library.

[0067] After completing the fault case feature extraction, the historical fault case data is saved in the case library according to the production system type.

[0068] Step 304: Train the RBF neural network model using the fault case library.

[0069] The features of the fault to be diagnosed are extracted to obtain the corresponding key feature attributes and their corresponding word vectors; based on this, the number of cluster center points is selected, and K-means clustering is performed on all relevant cases in the case library according to the word vectors to find cases with the same category as the case to be diagnosed, that is, similar cases; the fault features of each similar case are converted into word vectors and input into the trained RBF neural network for retrieval and reasoning, and the similarity with each similar case is output.

[0070] Step 305: extract features from the case to be diagnosed.

[0071] Step 306, case retrieval reasoning, to determine whether there are identical or similar cases to the case to be diagnosed. If there are similar cases, execute step 307; if there are identical cases, execute step 309.

[0072] Query past cases through key feature attributes, that is, search the fault case library. After vectorizing the extracted key feature attributes, select the number of cluster center points, perform K-means clustering on all cases in the fault case library according to the word vector, find cases with the same feature attributes as the case to be diagnosed, that is, similar cases, convert each fault feature of the same case into a word vector and input it into the trained RBF neural network, and output the similarity between the case to be diagnosed and each similar case in the output layer.

[0073] Step 307: Obtain suggested solutions from similar cases.

[0074] Step 308: Based on the suggested solution and in combination with the user's instruction, the suggested solution is modified to obtain a definite solution for the case to be diagnosed.

[0075] If a source case similar to the target case is retrieved, a suggested solution to the target case is obtained from the solution of the similar case, and the suggested solution is modified according to the actual situation. After that, a definite solution is obtained and the target case is saved as a new case in the case library.

[0076] Step 309: Obtain a definite solution from the same case.

[0077] If a source case identical to the target case is retrieved, a definite solution can be obtained and the target problem can be solved. Since the target case already has the same case, the target case can be discarded to avoid redundancy.

[0078] Step 310: Save the cases to be diagnosed and the determined solutions in the case library.

[0079] The present disclosure also provides a model training device for executing the model training method provided in any of the above embodiments. Figure 4 As shown, the device comprises: An acquisition unit 41 is used to acquire historical fault case data in a fault case library, wherein the historical fault case data includes fault description data and fault solutions; An obtaining unit 42, configured to obtain a fault prediction solution based on the fault description data and the neural network model to be trained; A calculation unit 43, configured to calculate a model loss based on the fault solution and the fault prediction solution; The optimization unit 44 is used to optimize the parameters of the neural network model using the model loss until the neural network model converges.

[0080] In some embodiments, the obtaining unit 42 is configured to: Extracting keywords from the fault description data according to predefined fault feature attributes to obtain fault feature keywords; Generating a fault feature vector of the fault feature keyword; The fault feature vector is input into the neural network model to obtain the fault prediction solution.

[0081] In some embodiments, the fault characteristic attribute includes at least one of the following: Operating environment; The fault belongs to; Fault phenomenon.

[0082] In some embodiments, the device is also used to: Obtaining a generation system type of the historical fault case data; The historical fault case data is stored in the fault case library according to the generation system type.

[0083] The present disclosure also provides a fault diagnosis device for executing the fault diagnosis method provided in any of the above embodiments. Figure 5 As shown, the device comprises: An acquisition module 51 is used to acquire target fault description data of a fault to be diagnosed; A prediction module 52, configured to predict a current fault prediction solution for the fault to be diagnosed by using a converged neural network model obtained based on the model training method in the aforementioned embodiment and the target fault description data; An acquisition module 53 is used to acquire, based on the current fault solution, a target historical fault diagnosis case having the highest similarity to the fault to be diagnosed from a historical fault diagnosis library; The determination module 54 is configured to use the fault diagnosis solution in the target historical fault diagnosis case as the fault solution for the fault to be diagnosed.

[0084] In some embodiments, the obtaining module 53 is used to: Extracting keywords from the target fault description data according to predefined fault feature attributes to obtain target fault feature keywords; Generate a target fault feature vector of the target fault feature keyword; Clustering the target fault feature vector and the fault feature vectors corresponding to the historical fault case data of the fault case library to obtain similar cases having the same characteristic attributes as the fault to be diagnosed; The similarity between the current fault solution and each of the similar cases is calculated, and the historical fault diagnosis case with the highest similarity is used as the target historical fault diagnosis case.

[0085] In some embodiments, the device is also used to: After taking the fault diagnosis solution in the target historical fault diagnosis case as the fault solution to the fault to be diagnosed, when it is determined that the fault description data of the target historical fault diagnosis case is not exactly the same as the target fault description data, the fault to be diagnosed and the fault solution to the fault to be diagnosed are updated to the fault history diagnosis library.

[0086] The disclosed embodiment also provides an electronic device to execute the above-mentioned model training method and fault diagnosis method. Figure 6 It shows a schematic diagram of an electronic device provided by some embodiments of the present disclosure. Figure 6 As shown, the electronic device 8 includes: a processor 800, a memory 801, a bus 802 and a communication interface 803, and the processor 800, the communication interface 803 and the memory 801 are connected via the bus 802; the memory 801 stores a computer program that can be run on the processor 800, and when the processor 800 runs the computer program, it executes the model training method and fault diagnosis method provided in any of the aforementioned embodiments of the present disclosure.

[0087] The memory 801 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk storage. The communication connection between the device network element and at least one other network element is realized through at least one communication interface 803 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0088] The bus 802 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory 801 is used to store programs, and the processor 800 executes the programs after receiving the execution instruction. The model training method and the fault diagnosis method disclosed in any implementation of the aforementioned embodiment of the present disclosure may be applied to the processor 800, or implemented by the processor 800.

[0089] The processor 800 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit in the processor 800 or the instruction in the form of software. The above processor 800 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The disclosed methods, steps and logic block diagrams in the embodiments of the present disclosure can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in conjunction with the embodiments of the present disclosure can be directly embodied as a hardware decoding processor to be executed, or the hardware and software modules in the decoding processor can be executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 801, and the processor 800 reads the information in the memory 801 and completes the steps of the above method in combination with its hardware.

[0090] The electronic device provided by the embodiments of the present disclosure and the model training method and fault diagnosis method provided by the embodiments of the present disclosure are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by them.

[0091] The present disclosure also provides a computer-readable storage medium corresponding to the model training method and fault diagnosis method provided in the above embodiments. Figure 7 The computer-readable storage medium shown is a CD 30 on which a computer program (ie, a program product) is stored. When the computer program is run by a processor, it will execute the model training method and fault diagnosis method provided by any of the aforementioned embodiments.

[0092] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical or magnetic storage media, which are not listed here one by one.

[0093] The computer-readable storage medium provided by the above-mentioned embodiments of the present disclosure and the model training method and fault diagnosis method provided by the embodiments of the present disclosure are based on the same inventive concept, and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0094] It should be noted that: In the description provided herein, a large number of specific details are described. However, it is understood that the embodiments of the present disclosure can be practiced without these specific details. In some instances, well-known structures and technologies are not shown in detail so as not to obscure the understanding of this description.

[0095] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of the exemplary embodiments of the present disclosure, the various features of the present disclosure are sometimes grouped together into a single embodiment, figure, or description thereof. However, the disclosed method should not be interpreted as reflecting the following schematic diagram: the claimed disclosure requires more features than the features explicitly recited in each claim. More specifically, as reflected in the claims below, the inventive aspects lie in less than all the features of the single embodiment disclosed above. Therefore, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, with each claim itself serving as a separate embodiment of the present disclosure.

[0096] In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of this disclosure and form different embodiments. For example, in the claims below, any one of the claimed embodiments may be used in any combination.

[0097] The above is only a preferred specific implementation of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed in the present disclosure should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A model training method, characterized in that: include: Acquire historical fault case data in a fault case library, wherein the historical fault case data includes fault description data and fault solutions; Obtaining a fault prediction solution based on the fault description data and the neural network model to be trained; calculating a model loss based on the fault solution and the fault prediction solution; The model loss is used to optimize the parameters of the neural network model until the neural network model converges.

2. The method according to claim 1, characterized in that Based on the fault description data and the neural network model to be trained, a fault prediction solution is obtained, including: Extracting keywords from the fault description data according to predefined fault feature attributes to obtain fault feature keywords; Generating a fault feature vector of the fault feature keyword; The fault feature vector is input into the neural network model to obtain the fault prediction solution.

3. The method according to claim 2, characterized in that The fault characteristic attribute includes at least one of the following: Operating environment; The fault belongs to; Fault phenomenon.

4. The method according to claim 1, characterized in that: Also includes: Obtaining a generation system type of the historical fault case data; The historical fault case data is stored in the fault case library according to the generation system type.

5. A fault diagnosis method, characterized in that: include: Obtain target fault description data of the fault to be diagnosed; Using a converged neural network model obtained by a model training method according to any one of claims 1 to 4 and the target fault description data, predicting a current fault prediction solution for the fault to be diagnosed; Based on the current fault solution, obtaining a target historical fault diagnosis case having the highest similarity to the fault to be diagnosed from a historical fault diagnosis library; The fault diagnosis solution in the target historical fault diagnosis case is used as the fault solution for the fault to be diagnosed.

6. The method according to claim 5, characterized in that Based on the current fault solution, obtaining a target historical fault diagnosis case with the highest similarity to the fault to be diagnosed from the historical fault diagnosis library includes: Extracting keywords from the target fault description data according to predefined fault feature attributes to obtain target fault feature keywords; Generate a target fault feature vector of the target fault feature keyword; Clustering the target fault feature vector and the fault feature vectors corresponding to the historical fault case data of the fault case library to obtain similar cases having the same characteristic attributes as the fault to be diagnosed; The similarity between the current fault solution and each of the similar cases is calculated, and the historical fault diagnosis case with the highest similarity is used as the target historical fault diagnosis case.

7. The method according to claim 5, characterized in that After taking the fault diagnosis solution in the target historical fault diagnosis case as the fault solution for the fault to be diagnosed, the method further includes: When it is determined that the fault description data of the target historical fault diagnosis case is not completely the same as the target fault description data, the fault to be diagnosed and the fault solution to the fault to be diagnosed are updated to the fault history diagnosis library.

8. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images